Research Of Colorful Image Segmentation Based On Improved FCM Algorithm

As the development of agricultural science and technology, machine vision technology has been applied in the agricultural equipment more widely than before. The accuracy of recognizing the crop target mainly depends on the accuracy of target segmentation. Most of existing flower image segmentation methods can only be limited on the partly specific flower image processing, which does not have universal applicability. Thus, segmenting flower goals of different types and colors from a variety of complex backgrounds will have great significance for improving the general applicability of image segmentation algorithm.Firstly, FCM(Fuzzy C-Means) algorithm has the advantages of simplicity, intuition and easy to implement. It is also highly sensitive to noise in image segmentation. Consequently, this paper presents an improved FCM image segmentation algorithm combined with the information of pixel neighborhood. Firstly, the algorithm segments image using FCM algorithm to produce preliminary segments results. It then uses the information of neighboring pixels to research the noise points in image. Finally, by amending membership of noise points, the impact of noise on the segmentation results can be eliminated.In order to solve the problem of BFO(Bacterial Foraging Optimization algorithm) in an iteration process caused by rapid population diversity loss which is easy to fall into local optima, this paper presents the combination of differential evolution ideas and BFO. Firstly, in the procedure of chemotaxis and reproduction of BFO, the paper has took differential policy to update bacteria position and ensure that diversity within population will not reduce too quickly with algorithm iteration process. Secondly, the paper has improved the updated method of the location of the bacteria in the algorithm by updating the position-by-dimensional rather than a way of updating information on all dimensions of a bacterium. The location updating method takes full advantage of the favorable information that every location change brings and greatly improves the efficiency and accuracy of the algorithm by finding the global optimum values.Finally, in order to solve the accurate extraction of various color flower goals in natural scenes, this paper presents a color image segmentation scheme of self-adaptive image data sources. In order to improve the performance of global search of standard FCM and accelerate the speed of algorithm convergence, this paper combines the improved bacterial foraging optimization algorithm and FCM. Meanwhile, this paper proposes a new fuzzy clustering evaluation function to assess the quality of an image segmentation results in different data sources so that the image segmentation method in this paper can achieve better segmentation results in different color flower objectives and complex natural backgrounds.In this paper, we have done a lot of experiments, and selected 10 representative images used as experimental objective. The results prove that the new scheme can adapt to changes in the target color and background factors. And, the image segmentation result is very close to expected human vision. Therefore the new solution has better universality than the traditional method.